Data Exploration

I’ve made some summary plots for each of the fitness traits we’ve measured over the course of the experiment - one plot for each trait, showing gardens~years and summarizing the sexuals (red) and apomicts (blue).

Establishment

First we’ve got the data from the establishment plots in each garden. These are just the means + SE for each mating system in each garden.

And here’s the means for each population in each garden

Survival

Size

Leaf length

Leaf Number

Flowering

Number of flowering individuals

Number of flower heads

Percent of surviving individuals that flowered.

Number of buds per individual planted over the course of the experiment. For this I simply summed up the number of buds produced by each mating system in each garden, and divided by the number of individuals planted at the start of the experiment. This plot shows almost exactly the same pattern as the plot from the aster model below…

Now the same thing but looking by population in each garden

This is somewhat crude, but gives a good visual exploration of what we’re hoping to look at. This takes (for each mating system) the mean establishment success in each garden * the number of buds per individual planted * the mean number of good seeds per bud (from source populations)

aster model

I’ve successfully made my first aster model! It’s pretty simple so far. It includes survival, flowering, and number of flowers as the fitness variables, and ‘mating system’ x ‘garden’ interaction as fixed effects (aster does have random effects capability, but the authors don’t like them and admonish us to only use them if absolutely necessary, and even then don’t put much faith in them…). I’ve ‘predicted’ lifetime fitness estimates and standard errors for each mating system in each garden and plotted them here:

I’ve also succesfully run an aster model with ‘ms x garden’ as fixed effects and ‘pop’ as a random effect, but there’s currently no way to ‘predict’ for random effects aster models because (as the authors caution) it is “very unclear” what prediction means for them and GLMMs in general. FWIW, the summary of the ‘reaster’ (random effects aster model) indicates that ‘pop’ does not have a strong impact on ‘ms x garden’ fixed effects - though I may be misinterpreting the model output.

I obviously still have a lot more to explore with aster models, but it’s a relief that I’ve been able to get them coded and running properly. Judging by the above figure, it seems quite clear that for sexuals, range limit ≠ niche limit! There’s also some indication of local adaptation / homesite advantage for the sexuals in the southern gardens. Other than that, there’s no obvious patterns emerging in the other garden regions…For example, in the MT gardens, sexuals had much higher fitness than apo’s in AA1, but much lower in AA2.